Last Updated on
Crypto-ML released R4 of our cryptocurrency machine learning models, which feature a breakthrough in anomaly detection. This technology leverages an entirely different approach than our previous models. It is expected to provide highly reliable trade signals to Crypto-ML members.
This post will cover a brief of the model and how to interpret the corresponding signal and graph on our Member Dashboard.
Anomaly Detection Overview
There are many approaches to machine learning. While our models R1 through R3 have achieved an exceptional Trade History, our machine learning team has been able to develop an entirely new approach that seeks to identify unusual behavior in the markets that tends to precede price changes.
These “unusual behaviors” include fundamental and technical data points, surveying a broad set of concepts. Ultimately, the model passes the data through a complex decision tree and decides whether or not it expects the markets to jump or drop.
This new model is very effective, and more importantly, extremely robust. This means its core concept seems to work across market conditions, across coins, and across time. Whereas other models need to constantly learn and adjust, this one seems to have tapped into the foundational pulse of the markets.
As an analogy, if this model was driving a car, it would do well on highways, on side roads, in parking lots, in other countries, and so on. It is broadly capable and highly robust. It “understands” the fundamentals of driving.
To learn how this machine learning model works, please read our extensive post Bitcoin Trading with Machine Learning Anomaly Detection. It also covers the concept of machine learning robustness.
As of this writing, the trade signals are issued for long Bitcoin trades, as Beta BTC.
- To receive trade alerts, include Continuous Alerts in your Notification Settings. You will receive BTC Beta alerts by email and mobile app notification.
- Additionally, Auto Trade can be set up for Beta BTC.
You can follow these trades manually with the Trader membership or automatically via the Auto Trader membership. See our free and paid membership plans on the Pricing Page.
Our prior trade alerts have been described as a “super technical indicator.” These models can be visualized with a form that oscillates over time. This information provides an indication of when a trade signal (buy or sell) may occur.
However, anomaly detection works differently. Rather than changing over time, it looks for unusual behavior, which may occur at any point without any warning.
Therefore our visualization shows a “pulse” when anomalous behavior is detected. These pulses help answer the question: Is it still okay to get in this trade?
For example, what if you missed the initial trade alert, but Bitcoin price will continue to climb for the next 30 days?
Instead of sitting on the sidelines, it would be great to know if it’s okay to still hop in and enjoy the ride.
On the BTC Anomaly Detector visualization, you’ll find pulses up and down. These pulses represent bullish and bearish anomalies in the market condition.
- Any pulse above zero represents a bullish expectation.
- Any pulse below zero represents a bearish expectation.
Each pulse indicates the expectation of future price.
If the position is closed:
- A bullish pulse will trigger a buy signal.
- A bearish pulse will indicate further downside is expected.
If the position is open:
- A bearish pulse will trigger a sell signal.
- A bullish pulse will indicate further upside is expected.
The best time to enter a trade is when the initial trade alert was generated. This is because our entire system is optimized toward this behavior: following the actual buy and sell signals as they are delivered. Deviating from this introduces risk.
However, if you missed the initial trade, yet still see recent bullish pulses, you may consider those data points as part of your decision to enter the trade late or not.
Bitcoin Anomaly Detection is just getting started. While we have performed extension testing and statistical analysis, the proof is in the pudding. We only post live trades delivered to our customers on the Crypto-ML Trade History page.
The model is off to a great start, but we’d like to see this at least a quarter’s worth of real-world results before we take it out of beta.
Please continue watching our Trade History page for the latest updates on performance.
And although it is currently in beta, our entire team has switched our Auto Trade selections to BTC Beta.
For access to the Bitcoin Anomaly Signals, the following memberships are needed:
- Signals only: Trader Membership (with free 30-day trial)
- Signals and Auto Trade: Auto Trader Membership (with free 30-day trial)
- View all of our memberships (free and paid): Pricing
If you have questions, comments, or suggestions, please let us know on the Community Forums!